Within the framework of the semantic mediation theory proposed by Osgood, predictions concerning mediated transfer were made. An experiment is reported in which the effects of three levels of semantic polarization of the mediating elements upon transfer in paired-associate learning were investigated in two mediation paradigms. The analysis of the data was divided into three main subanalyses. (1) Polarization and learnability . An analysis of the number of trials taken during learning stage I to reach a criterion of two perfect repetitions showed that as the polarization of the word increases, there is a decrease in the number of trials required to reach the criterion. (2) Meaningfulness and acquisition . The postassociation semantic profiles of the CVC items were found to shift significantly closer to the profile of the associated word from their own initial meaning profiles. Furthermore, a high degree of similarity was found between the terminal profiles of the CVC items and the profile for the associated word. (3.1) Mediation paradigm and transfer . Comparison of the transfer effects for simple chain and response-equivalence paradigms indicated that the former produces greater over-all learning. (3.2) Degree of second-stage learning and mediated transfer . Higher degrees of learning during the second-stage of either a simple chain or a response-equivalence paradigm lead to an increased mediated transfer effect. (3.3) Semantic content of the mediating element and mediated transfer . Increasing levels of polarization of the mediator resulted in significant increasing positive effects upon mediated transfer. It was concluded that semantic polarization of the meaningful element in CVC and WORD pairs has positive facilitative effects on rate and degree of learning and upon mediated transfer.